基於多層感知器與通道狀態資訊之室內定位

本文提出基於多層感知器(Multilayer Perceptron, MLP)和通道狀態資訊(Channel State Information, CSI)的室內定位方法。近年來快速發展的無線網路通訊技術，採用多天線和多通道將串列的資料利用並行的方式傳輸。然而，通道在資料傳輸過程中容易受到環境影響，而CSI根據通道的狀況記錄著完整的環境資訊。這些CSI的資訊隋著環境的不同呈現豐富的變化性，將有利於深度網路環境特徵的擷取。本文所提的方法與相較於傳統的方法擁有如下三個優勢。第一個優勢在於與傳統訊號強度定位方法相比，來自多通道的CSI具有更微細資訊。第二個優勢在於傳統採用手動分析以取得環境特徵。深度學習的多層感知器網路可擷取更微細環境特徵。第三個優勢在於傳統需在環境中架設多感測器，以便估計位置。我們採用的CSI來自無線網路傳輸，僅需架設無線網路存取點與配置有無線網卡的電腦即可。經由實驗的比較所提方法有較高的準確率。

關鍵字

室內定位；深度學習；機器學習；多層感知器；通道狀態資訊

並列摘要

An indoor positioning method based on Multilayer Perceptron (MLP) and Channel State Information (CSI) is proposed in this paper. In recent years, the rapid development of wireless network communication technology, the use of multiple antennas and multi-channel, the serial data transmission in parallel manner. However, the channel is vulnerable to environmental impact during data transmission, and CSI records complete environmental information according to the channel condition. These CSI information, rich in variations in the environment, will facilitate the retrieval of features from the deep network environment. The method presented in this paper has three advantages over the traditional method. The first advantage is that, compared with the traditional signal strength localization method, the multi-channel CSI has more fine information. The second advantage lies in the traditional use of manual analysis to obtain environmental features. Deep learning multilayer perceptron network for capturing finer environment features. The third advantage lies in the fact that traditional sensors need to be built in the environment to estimate the location. We use CSI from wireless network transmission, just set up wireless network access point and configure a wireless network card computer. The experimental results show that the method presented in this paper has more accurate than CNN-RSSI.

並列關鍵字

Indoor localization ； deep learning ； machine learning ； multilayer perceptron ； channel state information

國際替代計量

基於多層感知器與通道狀態資訊之室內定位

全文下載

主題瀏覽